Help needed, Cube disarmed during flight

Can somebody please help analyse these logfiles?

I’m totally lost. At the end of the flight, suddenly all kind of messages show up, the Cube disarms and we crash.

Is there anybody who can have a look please?

you’ve had a watchdog reset:

just a moment prior to that, something happened to your mag and EKF went bad. can’t tell if it caused or was caused by the wdg reset though. someone more familiar with wdg would have to look into this.

This is almost certainly caused by the i2c interrupt storm issue that was fixed in the 3.9.11 stable release. We’ve recently discovered that the QMC5883L can trigger this issue, and the log shows you have a QMC5883L compass. The symptoms match the interrupt storm that the QMC5883L can trigger. It is fixed fully in the 3.9.11 release, but I know that isn’t of much comfort to you now.
It should have started logging again after the reset, but it doesn’t seem to have done that. I will test locally to ensure that is working.
You could have armed again after the reset as it would have bypassed all arming checks, but given the amount of time you have when this happens it is hard to react fast enough, even if you know to do this.

What gps was fitted?

Thank you very much for the help. At least we know it wasn’t something we did wrong :smirk:. And we were flying when you released the fix… Such bad luck.

The Here GPS. We always buy the bundle from jdrones.

Can you please send a picture of the GPS?
Our here gps is not fitted with that compass, so if that compass is being detected, then there may be a different issue

Did you have a secondary GPS?

We use other compass. The compass we use is

No, we not use secondary gps.

ok, @jani_hirvinen, which mag is on that? Is it the QMC5883L?

The here is not fitted with this magnetometer

With all due respect, I am satisfied with the answer from @tridge. Of course it is always a tragedy to loose an airplane due to some bug.

But for me, this is fully acceptable. This is open source software, and thanks to the hard work of so many people, it is freely available for everyone.

I consider a crash like ours a valuable contribution to the community. It will make the software better, more reliable and stable for the whole community.

I appreciate your efforts to help investigating and I hope it will leed to even more understanding of the failure.

1 Like

Why would that airspeed/compass board claim to have a HMC5983 mag yet the device ID identifies it as a QMC5883L would be my question.

1 Like

The reason that we continue to investigate, is that according to the components you say you had fitted, the detected Magnotometer is not fitted to your vehicle, so we need to find out if that is correct.

If the suspect magnetometer was fitted, then Tridge’s solution will be satisfactory.

If the suspect magnetometer was NOT fitted, then that would indicate a secondary issue that needs to be cleared up.

Can you take a close up picture of your magnetometer on the airspeed sensor?

No problem. Please find attached:

https://drive.google.com/a/top-engcorp.com/file/d/11gmgJ5XEBw_82xmI65VzYLj8L4uNaaiB/view?usp=drivesdk

https://drive.google.com/a/top-engcorp.com/file/d/11b3kEIxCRatx8eny4ULCetsZXicA4xTx/view?usp=drivesdk

Depending on the age of the airspeed/mag board it’s either HMC5983 or QMC5883.

Awesome, so we are not chasing another bug! Thanks Jani!

You were absolutely right. There is a second logfile:

00000005.BIN (676 KB)

Thanks! That is consistent with the conclusions above.
I’d also recommend that you not use that specific compass again as it was clearly acting up. The new firmware will prevent that misbehaviour from causing the watchdog reset you saw, but it may still get bad data. In the few seconds before the lockup it gave some odd spikes that it really shouldn’t be giving